So there's these enhancements that allow you to "deterministically map MVPN
groups to specific MDT groups".
I gather the following, from playing with it:
- you still only get one "mdt data <net> <mask>", only now you get to say "list
<acl>"
- otherwise it falls through to the default
- to the extent it's possible, you get a sequential-match, e.g.
ip access-list standard my-groups
permit 233.254.99.128 0.0.0.63
ip vrf BLAH
mdt default 233.1.1.1
mdt data 233.88.88.64 0.0.0.63
you get 233.254.99.129 -> 233.88.88.65
233.254.99.134 -> 233.88.88.70
ad nauseum
Are there other feasible incantations?
The reason I ask is because, well, I use a deterministic method for allocating
the MDT groups, so I know that VRF BLAH -> default 225.1.1.A data
225.1.A.<x-y>, and a lot of the input groups are x.y.z.<0-255>, so it'd be damn
handy in debug-panic to not have to continually do a "sho ip pim vrf BLAH mdt
<err am I sending or receiving on this switch anyway?> <err so which group in
this list is what?>"...
Is there any really good reason to run a bunch of MDT data groups besides to
allow breaking down the groups in the MVPN so as to allow multicast/SSM in
global space to optimize what gets sent where?
Is there a downside to letting MVPN "splay" all the groups out, other than
possibly running out of TCAM at some N aggregate number of groups, noting that
each MVPN group really takes up two slots in
MFIB/<whatever-it's-called-on-non-sup2t>? (Well, that and having "show ip
mroute" scroll for a really long time, I suppose.)
(I figure I'm running around 400 "real" groups input into the mesh at various
points, ranging from 10pps to 50kpps per group. 6500s all, though ME3600s will
probably enter the picture.)
_______________________________________________
cisco-nsp mailing list [email protected]
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/
